The issue of the design effect in water, sanitation and hygiene studies

摘要

Cluster sampling is commonly used in water, sanitation and hygiene (WASH) surveys, as in the Multiple Indicator Cluster Survey developed by the United Nations Children's Fund (UNICEF) for the assessment of development-related goals. In cluster survey techniques, despite a good approximation of the design effect is essential for efficient sample size determination and for obtaining accurate precision of survey estimates; the assessment of this parameter has often been overlooked. This study computes the design effects for three core WASH outcomes at two different administrative scales. We use the database of a Kenyan case study for this purpose. We show that design effects differ greatly, and large differences have been found for different variables, different regional setting, and different scale of analysis. We recommend that survey planners should keep in mind such differences when defining the objectives of the survey and the required precision of survey estimates.